Anti-interference line-of-sight stabilization control method for dynamic target tracking

Abstract

An anti-interference line-of-sight stabilization control method for dynamic target tracking belongs to the field of visual servoing and photoelectric tracking technology. This invention addresses the problem that existing line-of-sight stabilization systems cannot effectively handle time-varying and unknown target maneuvers. It includes: mounting a camera on the line-of-sight stabilization device to acquire target images and obtain the image miss distance; using a UDE-based outer-loop programmable logic controller to obtain an interference estimation signal; using an MPC-based outer-loop visual servo controller to calculate the outer-loop control signal; a UDE-based inner-loop attitude controller to obtain a control signal for the line-of-sight stabilization device based on the outer-loop control signal and the current attitude of the line-of-sight stabilization device; and the actuator of the line-of-sight stabilization device adjusting its attitude according to the control signal, ensuring that the image miss distance of the target position coordinates in the target image is within a set threshold. This invention is applicable to scenarios such as UAV monitoring, UAV countermeasures, and low-altitude security.

Description

Technical Field

[0001] This invention relates to an anti-interference line-of-sight stabilization control method for dynamic target tracking, belonging to the field of visual servoing and photoelectric tracking technology. Background Technology

[0002] In recent years, with the rapid popularization of low-altitude small aircraft technology, drones have played an important role in logistics delivery, emergency rescue, and other fields, but have also brought safety hazards, necessitating high-performance low-altitude target tracking technology as a core support for monitoring and countermeasures. Line-of-sight (LAS) stabilization systems are key equipment for achieving low-altitude target tracking. They capture target information through photoelectric detectors and drive actuators to adjust attitude, completing stable tracking and aiming at the target. Due to their high detection accuracy and strong concealment, LAS play a central role in target tracking missions. However, existing LAS stabilization control technology has significant shortcomings:

[0003] 1. Insufficient ability to deal with maneuvering targets: Most control methods are based on the assumption that the target moves at a constant speed or with uniform acceleration, which makes it difficult to effectively deal with the time-varying and unknown maneuvers of the target, resulting in estimation model mismatch, increased tracking error, or even target loss.

[0004] 2. Limited anti-interference capability: The tracking system itself has model uncertainties such as friction, load imbalance, and parameter perturbation, and is also subject to external environmental disturbances. Existing control strategies often lack effective online estimation and compensation methods for such multi-source and complex disturbances.

[0005] 3. Disconnection between perception and control links: Most studies focus on individual trajectory tracking control or mechanism modeling, lacking integrated design and verification of "target detection-information processing-servo control", which leads to a decline in system performance during actual closed-loop operation.

[0006] 4. Performance Degradation Due to Bandwidth Bottleneck: The bandwidth of the outer loop visual servo loop is limited by the attitude response capability of the inner loop and the sensor update rate. Under these conditions, traditional low-pass filter-based interference observers will exhibit significant phase lag, severely affecting the estimation and compensation timeliness of rapidly changing interference, making it difficult to ensure that the target remains within the field of view.

[0007] Therefore, there is an urgent need in this field for a new line-of-sight stabilization control scheme that can comprehensively address the uncertainties of target maneuvering and system model, and achieve anti-interference under actual bandwidth constraints. Summary of the Invention

[0008] To address the problem that existing line-of-sight stabilization systems cannot effectively handle time-varying and unknown maneuvers of targets, this invention provides an anti-interference line-of-sight stabilization control method for dynamic target tracking.

[0009] An anti-interference line-of-sight stabilization control method for dynamic target tracking according to the present invention includes:

[0010] A camera is mounted on the line-of-sight stabilization device to acquire target images; the target images are processed to obtain the target position coordinates in the camera plane coordinate system, and the image miss distance of the target position coordinates is obtained.

[0011] An outer-loop programmable logic controller based on UDE is used to calculate the interference estimation signal based on the image miss distance and the current attitude of the line-of-sight stabilization device.

[0012] An MPC-based outer-loop vision servo controller calculates the outer-loop control signal based on the image miss distance, interference estimation signal, and the current attitude of the vision axis stabilization device.

[0013] The UDE-based inner-loop attitude controller obtains the control signal for the line-of-sight stabilization device based on the outer-loop control signal and the current attitude of the line-of-sight stabilization device.

[0014] The actuator of the line-of-sight stabilization device adjusts the attitude of the line-of-sight stabilization device according to the control signal to the line-of-sight stabilization device, so that the image miss distance of the target position coordinates in the target image is within a set threshold.

[0015] The anti-interference line-of-sight stabilization control method for dynamic target tracking according to the present invention includes the following method for obtaining the image miss distance:

[0016] A deep learning-based target detection algorithm is used to process the target image to obtain the target position coordinates in the camera plane coordinate system. The off-target amount is calculated based on the target position coordinates and the camera principal point center coordinates.

[0017] According to the anti-interference line-of-sight stabilization control method for dynamic target tracking of the present invention, the target position coordinates are set as follows: In the formula The horizontal pixel coordinates of the target location. The vertical pixel coordinates of the target location; the coordinates of the camera principal point center. In the formula The horizontal pixel coordinates of the camera's principal point center. The vertical pixel coordinates of the camera's principal point center;

[0018] Then the off-target amount of the image for:

[0019] ,

[0020] In the formula This represents the horizontal pixel miss distance. This refers to the vertical pixel miss distance.

[0021] Image off-target amount Satisfy visibility constraints :

[0022] ,

[0023] In the formula For the set of real numbers, The coordinates of the maximum position within the camera's field of view:

[0024] ,

[0025] In the formula The maximum coordinates of the horizontal pixels. This represents the maximum coordinates of the vertical pixels.

[0026] The interference estimation signal is obtained using the anti-interference line-of-sight stabilization control method for dynamic target tracking according to the present invention.

[0027] The outer-loop programmable logic controller based on UDE adopts a servo control form with a binary structure:

[0028] ,

[0029] In the formula The vector form of the angular velocity of the line-of-sight stabilization device. The nominal angular velocity control quantity is calculated by the MPC-based outer-loop vision servo controller. Let be the Jacobian matrix representing the rate of change from angular velocity to miss distance. For interference estimation signal;

[0030] ,

[0031] In the formula To estimate the horizontal component of the signal to mitigate interference, To estimate the vertical component of the signal to mitigate interference;

[0032] according to ,

[0033] In the formula The pitch angle of the line-of-sight stabilizing device. The uncertainty term is caused by the unknown target velocity. The target velocity in the coordinate system of the line-of-sight stabilized device. The X-axis component represents the position of the target in the coordinate system of the line-of-sight stabilized device.

[0034] ,

[0035] In the formula The horizontal component of the uncertainty term caused by the unknown target velocity. The vertical component of the uncertainty term caused by the unknown target velocity;

[0036] get: ,

[0037] In the formula For the virtual input signals of the UDE-based outer-loop programmable logic controller:

[0038] ,

[0039] In the formula for Horizontal components, for The vertical component;

[0040] ;

[0041] right After performing low-pass filtering and phase lead correction, we obtain:

[0042]

[0043] In the formula To estimate the frequency domain form of the interference signal, For frequency domain operators, For the filter transfer function, For the Laplace operator, For time, For phase lead correction parameter one, This is the second phase lead correction parameter. These are the parameters of the low-pass filter;

[0044] Taking the inverse Laplace transform of the above equation, we get:

[0045] ,

[0046] In the formula For convolution operators, For integration variables in the time domain;

[0047] Similarly, we get:

[0048] .

[0049] The anti-interference line-of-sight stabilization control method for dynamic target tracking according to the present invention includes the cost function of the UDE-based inner-loop attitude controller. Designed as follows:

[0050] ,

[0051] In the formula This is based on the nominally predicted future miss rate. For future control signals to be optimized, To predict duration, To predict the quadratic cost index that considers miss convergence and energy consumption in the time domain, For the integral variable in the cost function, This is an indicator of terminal error cost.

[0052] According to the anti-interference line-of-sight stabilization control method for dynamic target tracking of the present invention, in Time interval with respect to cost function The optimized control is as follows:

[0053] ,

[0054] In the formula for Predicted pitch angle at any given time.

[0055] According to the anti-interference line-of-sight stabilization control method for dynamic target tracking of the present invention, the outer loop control signal includes an angular velocity reference command. and angle reference instructions :

[0056] ,

[0057] In the formula The angular velocity command is obtained from the MPC-based outer-loop vision servo controller. The angular velocity command is obtained from the outer-loop programmable logic controller based on UDE. To The optimal nominal control signal obtained through optimization;

[0058] Integrating the above equation yields the angle reference command. :

[0059] ;

[0060] The output outer loop control signal is then... .

[0061] According to the anti-interference line-of-sight stabilization control method for dynamic target tracking of the present invention, the UDE-based inner-loop attitude controller first converts the outer-loop control signal... Decomposed into pitch channel attitude commands and yaw channel attitude commands In the formula For pitch angle command, This is the yaw angle command;

[0062] The nominal control input for the pitch channel of the line-of-sight stabilization device. for:

[0063] ,

[0064] In the formula The pitch channel inertia coefficient, For pitch channel proportional feedback gain. For the pitch channel differential feedback gain, ; For pitch angle error, ; These are the damping parameters for the pitch channel;

[0065] The nominal control input for the yaw channel of the line-of-sight stabilization device. for:

[0066] ,

[0067] In the formula The inertia coefficient of the yaw channel. This is the proportional feedback gain for the yaw channel. For the differential feedback gain of the yaw channel, ; For yaw angle error, , The yaw angle of the line-of-sight stabilizing device; These are the damping parameters for the yaw channel.

[0068] According to the anti-interference line-of-sight stabilization control method for dynamic target tracking of the present invention, the UDE-based inner-loop attitude controller is also used to calculate the pitch channel interference estimate. And yaw channel interference estimates :

[0069] ,

[0070] ,

[0071] In the formula For pitch channel UDE parameters, ; For yaw channel UDE parameters, .

[0072] According to the anti-interference line-of-sight stabilization control method for dynamic target tracking of the present invention, the control signal for the line-of-sight stabilization device output by the UDE-based inner-loop attitude controller includes a pitch channel composite control signal. Combined control signal with yaw channel :

[0073] ;

[0074] .

[0075] The beneficial effects of this invention are as follows: The method of this invention, applied to automatic control, visual servoing, and photoelectric precision tracking technologies, solves the problems of poor tracking accuracy and weak robustness in existing line-of-sight stabilization systems when facing unknown target maneuvers and uncertainties in their own models. This invention employs a dual-loop robust control architecture. The outer loop is an image-based servo loop, innovatively designed with a disturbance observer featuring phase lead compensation, effectively overcoming phase lag under bandwidth constraints and achieving accurate estimation of target motion uncertainties. Combined with robust model predictive control, it generates optimal attitude commands online under strict field-of-sight constraints that ensure the target is not lost. The inner loop is an attitude tracking loop, employing a disturbance estimator to compensate for internal model uncertainties and external disturbances, achieving fast and accurate tracking of commands. This invention significantly improves the tracking accuracy and environmental adaptability of line-of-sight stabilization systems for non-cooperative maneuvering targets, making it suitable for scenarios such as UAV monitoring, UAV countermeasures, and low-altitude security, and possessing significant engineering application value. Attached Figure Description

[0076] Figure 1 This is a schematic diagram of the coordinate system definition and tracking scenario in the anti-interference line-of-sight stabilization control method for dynamic target tracking described in this invention;

[0077] Figure 2 This is a control block diagram of the anti-interference line-of-sight stabilization control method for dynamic target tracking described in this invention. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] Specific Implementation Method 1: Combination Figure 1 and Figure 2 As shown, the present invention provides an anti-interference line-of-sight stabilization control method for dynamic target tracking, comprising:

[0080] A camera is mounted on the line-of-sight stabilization device to acquire target images; the target images are processed to obtain the target position coordinates in the camera plane coordinate system, and the image miss distance of the target position coordinates is obtained.

[0081] An outer-loop programmable logic controller based on UDE is used to calculate the interference estimation signal based on the image miss distance and the current attitude of the line-of-sight stabilization device.

[0082] An MPC-based outer-loop vision servo controller calculates the outer-loop control signal based on the image miss distance, interference estimation signal, and the current attitude of the vision axis stabilization device.

[0083] The UDE-based inner-loop attitude controller obtains the control signal for the line-of-sight stabilization device based on the outer-loop control signal and the current attitude of the line-of-sight stabilization device.

[0084] The actuator of the line-of-sight stabilization device adjusts the attitude of the line-of-sight stabilization device according to the control signal to the line-of-sight stabilization device, so that the image miss distance of the target position coordinates in the target image is within a set threshold.

[0085] The overall execution logic of this implementation follows the concept of hierarchical control, including image detection, outer loop visual servo control, and inner loop attitude tracking. The system block diagram is as follows: Figure 2 As shown.

[0086] Image detection involves acquiring target images using a camera mounted on a line-of-sight stabilization device, identifying targets using a deep learning-based target detection algorithm (You Only Look Once, YOLO), extracting the target's center pixel coordinates, and calculating the image's miss distance (horizontal and vertical deviation) by comparing it with the camera's principal point center coordinates.

[0087] External loop vision servo control includes:

[0088] Based on the image miss distance and the target motion model in the line-of-sight frame, an uncertainty and disturbance estimator with phase-lead compensation (UDE-PLC) is used to estimate the uncertainty and disturbance caused by the unknown maneuvering speed of the target online. The phase lag problem caused by the limited outer loop bandwidth is also mitigated by the lead compensation stage.

[0089] Based on the constructed visual kinematics model, a robust model predictive control (MPC) is designed. This controller, based on predictions of the system state over a future period, optimizes control performance while strictly ensuring that the image miss distance does not exceed the field of view (i.e., visibility constraint). The estimated values ​​from the disturbance observer are feedforwarded and compensated, and combined with the optimal control output from the MPC, to generate the desired frame angular velocity command.

[0090] Integrate the angular velocity command to obtain the frame angle command; obtain the angular acceleration command through the differential tracker.

[0091] Inner loop attitude tracking includes:

[0092] Receives angle, angular velocity, and angular acceleration commands generated by the external circuit.

[0093] The uncertainty and disturbance estimator is used to quickly estimate the uncertainty of the inner loop.

[0094] Design a composite controller that combines feedforward, feedback, and disturbance compensation to drive actuators (such as motors) to quickly and accurately track attitude commands given by the external loop.

[0095] System Modeling and Problem Construction: Establish a dynamic model of the line-of-sight stabilization system and a visual kinematic model based on image features. Clearly characterize the impact of unknown target maneuvers as external loop uncertainty, and characterize the impact of internal parameter changes, friction, etc. as internal loop uncertainty. At the same time, define the field-of-sight constraints that must be satisfied during the tracking process.

[0096] Scene and coordinate system modeling:

[0097] This implementation considers a scenario where an electro-optical tracking system is tracking a single low-altitude flying target. The scenario and coordinate system are defined as follows: Figure 1 As shown: Includes inertial frame visual axis system and pixel system The target inertial frame position is... The speed is The attitude of the optical television axis system is determined by the pitch angle. and azimuth It is represented in vector form as The angular velocity in vector form is The inertial frame and the line-of-sight frame share the same origin, and the rotation sequence is "azimuth first, then pitch." Therefore, the target's position within the line-of-sight frame... It can be represented as:

[0098] ,

[0099] The velocity of the target in the line-of-sight frame is:

[0100] ,

[0101] in Representing the target velocity through the rotation matrix Transformed to the form under the line of sight, where The antisymmetric matrix representing angular velocity. To characterize the angular velocity of the tracking system in the line-of-sight frame, namely:

[0102] .

[0103] Furthermore, kinematic modeling of image miss distance:

[0104] Methods for obtaining the off-target amount in an image include:

[0105] A deep learning-based target detection algorithm is used to process the target image to obtain the target position coordinates in the camera plane coordinate system. The off-target amount is calculated based on the target position coordinates and the camera principal point center coordinates.

[0106] Set the target position coordinates as In the formula The horizontal pixel coordinates of the target location. The vertical pixel coordinates of the target location; the coordinates of the camera principal point center. In the formula The horizontal pixel coordinates of the camera's principal point center. The vertical pixel coordinates of the camera's principal point center;

[0107] Then the off-target amount of the image for:

[0108] ,

[0109] In the formula This represents the horizontal pixel miss distance. This refers to the vertical pixel miss distance.

[0110] The camera's focal length is Based on the pinhole imaging principle of a camera, the relationship between the miss distance and the projection of the target in the line of sight is as follows:

[0111] ,

[0112] in and Let the focal lengths be the horizontal and vertical focal lengths, respectively. Taking the derivative of the above equation and using formulas for target velocity, we can obtain the following model:

[0113] ,

[0114] ;

[0115] in It is the Jacobian matrix of the rate of change from angular velocity to miss distance. This represents the uncertainty caused by the unknown target velocity. The field of view of commonly used photoelectric cameras is often limited. To ensure stable capture of target features, the image miss distance is... Satisfy visibility constraints :

[0116] ,

[0117] In the formula For the set of real numbers, The coordinates of the maximum position within the camera's field of view:

[0118] ,

[0119] In the formula The maximum coordinates of the horizontal pixels. This represents the maximum coordinates of the vertical pixels.

[0120] Tracking system dynamics modeling:

[0121] The dynamics of pitch and yaw can be described in the following Lagrangian form:

[0122] ,

[0123] in The driving torque generated by the motor, External input interference, Represents the inertia matrix. Represents the damping matrix. These are the inertia coefficients of the two channels. For damping parameters, Let represent the inertial torque and Coriolis torque caused by the base motion. Considering only the target tracking scenario of the fixed base tracking system, and treating factors such as friction and load gravitational torque as model uncertainties, and lumped together with external disturbances, the controlled object model can be obtained as follows:

[0124] ,

[0125] in This indicates aggregate uncertainty.

[0126] This implementation uses YOLO-based target detection:

[0127] To achieve stable and accurate perception of maneuvering targets and provide reliable input signals for the servo control system, this implementation method employs a target detection scheme based on the YOLO series deep learning models. The specific execution steps and technical details of this module are as follows:

[0128] Step 1: Scene Modeling and Data Acquisition. A virtual scene is constructed using a 3D simulation platform, incorporating various low-altitude backgrounds (urban, suburban, cloud / haze interference, etc.) and typical UAV targets. By adjusting the focal length, field of view, and target distance (near, medium, and long range) of the electro-optical camera, a large number of first-person perspective tracking image sequences are systematically generated.

[0129] Step 2: Data annotation. Use image annotation tools to accurately annotate the targets in the acquired images, generating a standard dataset containing target bounding boxes and category information.

[0130] Step 3: Model Selection and Optimized Training. YOLOv11 is preferred as the base network architecture. This architecture offers a good balance between detection speed and accuracy, and its backbone, neck network, and detection head design effectively extract multi-scale features, making it suitable for moving target detection scenarios at different scales. Furthermore, CUDA and cuDNN are configured based on PyTorch or TensorFlow deep learning frameworks to accelerate training. A stochastic gradient descent optimizer is used to train the model, and automatic mixed-precision training is enabled to save GPU memory and improve training speed.

[0131] Step 4: Model Deployment and Real-Time Inference. TensorRT is used to quantize and optimize the model, significantly improving inference speed on NVIDIA hardware. Based on the YOLO model, target locations are detected in real-time, bounding boxes are labeled, and the center coordinates of each bounding box are calculated. .

[0132] Step 5: Calculate the miss distance. Set the target center point coordinates... With the principal point coordinates of the image The difference is used to obtain the off-target amount in the image. This off-target amount This is the direct input to the external loop of the vision servo system.

[0133] Outer loop vision servo control based on UDE-PLC and MPC:

[0134] The method for obtaining the interference estimation signal is as follows:

[0135] The outer-loop programmable logic controller based on UDE adopts a servo control form with a binary structure:

[0136] ,

[0137] In the formula The vector form of the angular velocity of the line-of-sight stabilization device. The nominal angular velocity control quantity is calculated by the MPC-based outer-loop vision servo controller. Let be the Jacobian matrix representing the rate of change from angular velocity to miss distance. For interference estimation signal;

[0138] ,

[0139] In the formula To estimate the horizontal component of the signal to mitigate interference, To estimate the vertical component of the signal to mitigate interference;

[0140] according to ,

[0141] In the formula The pitch angle of the line-of-sight stabilizing device. The uncertainty term is caused by the unknown target velocity. The target velocity in the coordinate system of the line-of-sight stabilized device. The X-axis component represents the position of the target in the coordinate system of the line-of-sight stabilized device.

[0142] ,

[0143] In the formula The horizontal component of the uncertainty term caused by the unknown target velocity. The vertical component of the uncertainty term caused by the unknown target velocity;

[0144] get: ,

[0145] In the formula For the virtual input signals of the UDE-based outer-loop programmable logic controller:

[0146] ,

[0147] In the formula for Horizontal components, for The vertical component;

[0148] ;

[0149] A composite structure of "low-pass filtering + phase lead correction" is adopted for... After performing low-pass filtering and phase lead correction, we obtain:

[0150]

[0151] In the formula To estimate the frequency domain form of the interference signal, For frequency domain operators, For the filter transfer function, For the Laplace operator, For time, For phase lead correction parameter one, This is the second phase lead correction parameter. These are the parameters of the low-pass filter; ;

[0152] Taking the inverse Laplace transform of the above equation, we get:

[0153] ,

[0154] In the formula For convolution operators, For integration variables in the time domain;

[0155] Similarly, we get:

[0156] .

[0157] ,

[0158] In the formula It is a standard first-order low-pass filter. This is an advanced correction step;

[0159] Furthermore, a vision servo controller based on MPC is designed. The MPC controller uses the previously derived miss kinematic model to predict the future system state in the time domain. While ensuring state constraints, it provides online optimized control signals to achieve miss convergence. The cost function of the UDE-based inner-loop attitude controller is... Designed as follows:

[0160] ,

[0161] In the formula This is based on the nominally predicted future miss rate. For future control signals to be optimized, To predict duration, To predict the quadratic cost index that considers miss convergence and energy consumption in the time domain, For the integral variable in the cost function, This is an indicator of terminal error cost.

[0162] , ;

[0163] For a given vector and diagonal array Its norm is according to Defined in this way, and The cost matrix in the formula is defined as follows: , and Designed based on cost function The time-based optimization control problem is as follows:

[0164] exist Time interval with respect to cost function The optimized control is as follows:

[0165] ,

[0166] In the formula for Predicted pitch angle at any given time.

[0167] The above constraints ensure that the target remains within the field of view at all times. The first control action obtained by solving the above optimization problem at each step is used as the output of MPC.

[0168] Composite control signals and attitude trajectory generation. Combining the outputs of the UDE-PLC and MPC, the outer loop control signal includes angular velocity reference commands. and angle reference instructions :

[0169] ,

[0170] In the formula The angular velocity command is obtained from the MPC-based outer-loop vision servo controller. The angular velocity command is obtained from the outer-loop programmable logic controller based on UDE. To The optimal nominal control signal obtained through optimization;

[0171] Integrating the above equation yields the angle reference command. :

[0172] ;

[0173] right Differentiation yields the angular acceleration signal. The output outer loop control signal is then... .

[0174] UDE-based inner-loop attitude tracking control: Receives and parses attitude command signals generated by the outer loop. The inner loop employs a dual-channel control mode.

[0175] The UDE-based inner-loop attitude controller first processes the outer-loop control signal... Decomposed into pitch channel attitude commands and yaw channel attitude commands In the formula For pitch angle command, This is the yaw angle command;

[0176] For the pitch channel, the nominal control employs a feedforward + feedback + model cancellation approach; therefore, the nominal control input for the pitch channel of the line-of-sight stabilization device... for:

[0177] ,

[0178] In the formula The pitch channel inertia coefficient, For pitch channel proportional feedback gain. For the pitch channel differential feedback gain, ; For pitch angle error, ; These are the damping parameters for the pitch channel; This refers to the pitch angular velocity error. Similarly, the yaw channel is also designed with this control method.

[0179] The nominal control input for the yaw channel of the line-of-sight stabilization device. for:

[0180] ,

[0181] In the formula The inertia coefficient of the yaw channel. This is the proportional feedback gain for the yaw channel. For the differential feedback gain of the yaw channel, ; For yaw angle error, , The yaw angle of the line-of-sight stabilizing device; These are the damping parameters for the yaw channel.

[0182] The UDE-based inner-loop attitude controller is also used to calculate pitch channel disturbance estimates. And yaw channel interference estimates :

[0183] ,

[0184] ,

[0185] In the formula For pitch channel UDE parameters, ; For yaw channel UDE parameters, .

[0186] The control signals output by the UDE-based inner-loop attitude controller for the line-of-sight stabilization device include a composite control signal from the pitch channel. Combined control signal with yaw channel :

[0187] ;

[0188] .

[0189] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. An anti-interference line-of-sight stabilization control method for dynamic target tracking, characterized in that, include: A camera is mounted on the line-of-sight stabilization device to acquire target images; The target image is processed to obtain the target position coordinates in the camera plane coordinate system, and the image miss distance of the target position coordinates is obtained. An outer-loop programmable logic controller based on UDE is used to calculate the interference estimation signal based on the image miss distance and the current attitude of the line-of-sight stabilization device. An MPC-based outer-loop vision servo controller calculates the outer-loop control signal based on the image miss distance, interference estimation signal, and the current attitude of the vision axis stabilization device. The UDE-based inner-loop attitude controller obtains the control signal for the line-of-sight stabilization device based on the outer-loop control signal and the current attitude of the line-of-sight stabilization device. The actuator of the line-of-sight stabilization device adjusts the attitude of the line-of-sight stabilization device according to the control signal to the line-of-sight stabilization device, so that the image miss distance of the target position coordinates in the target image is within a set threshold.

2. The anti-interference line-of-sight stabilization control method for dynamic target tracking according to claim 1, characterized in that, Methods for obtaining the off-target amount in an image include: A deep learning-based target detection algorithm is used to process the target image to obtain the target position coordinates in the camera plane coordinate system. The off-target amount is calculated based on the target position coordinates and the camera principal point center coordinates.

3. The anti-interference line-of-sight stabilization control method for dynamic target tracking according to claim 2, characterized in that, Set the target position coordinates as In the formula The horizontal pixel coordinates of the target location. The vertical pixel coordinates of the target location; the coordinates of the camera principal point center. In the formula The horizontal pixel coordinates of the camera's principal point center. The vertical pixel coordinates of the camera's principal point center; Then the off-target amount of the image for: ， In the formula This refers to the horizontal pixel miss distance. This refers to the vertical pixel miss distance. Image off-target amount Satisfy visibility constraints : ， In the formula For the set of real numbers, The coordinates of the maximum position within the camera's field of view: ， In the formula The maximum coordinates of the horizontal pixels. This represents the maximum coordinates of the vertical pixels.

4. The anti-interference line-of-sight stabilization control method for dynamic target tracking according to claim 3, characterized in that, The method for obtaining the interference estimation signal is as follows: The outer-loop programmable logic controller based on UDE adopts a servo control form with a binary structure: ， In the formula The vector form of the angular velocity of the line-of-sight stabilization device. The nominal angular velocity control quantity is calculated by the MPC-based outer-loop vision servo controller. Let be the Jacobian matrix representing the rate of change from angular velocity to miss distance. For interference estimation signal; ， In the formula To estimate the horizontal component of the signal to mitigate interference, To estimate the vertical component of the signal to mitigate interference; according to , In the formula The pitch angle of the line-of-sight stabilizing device. The uncertainty term is caused by the unknown target velocity. The target velocity in the coordinate system of the line-of-sight stabilized device. The X-axis component represents the position of the target in the coordinate system of the line-of-sight stabilized device. ， In the formula The horizontal component of the uncertainty term caused by the unknown target velocity. The vertical component of the uncertainty term caused by the unknown target velocity; get: , In the formula For the virtual input signals of the UDE-based outer-loop programmable logic controller: ， In the formula for Horizontal components, for The vertical component; ； right After performing low-pass filtering and phase lead correction, we obtain: In the formula To estimate the frequency domain form of the interference signal, For frequency domain operators, For the filter transfer function, For the Laplace operator, For time, For phase lead correction parameter one, This is the second phase lead correction parameter. These are the parameters of the low-pass filter; Taking the inverse Laplace transform of the above equation, we get: ， In the formula For convolution operators, For integration variables in the time domain; Similarly, we can obtain: 。 5. The anti-interference line-of-sight stabilization control method for dynamic target tracking according to claim 4, characterized in that, Cost function of UDE-based inner-loop attitude controller Designed as follows: ， In the formula This is based on the nominally predicted future miss rate. For future control signals to be optimized, To predict duration, To predict the quadratic cost index that considers miss convergence and energy consumption in the time domain, For the integral variable in the cost function, This is an indicator of terminal error cost.

6. The anti-interference line-of-sight stabilization control method for dynamic target tracking according to claim 5, characterized in that, exist Time interval with respect to cost function The optimized control is as follows: ， In the formula for Predicted pitch angle at any given time.

7. The anti-interference line-of-sight stabilization control method for dynamic target tracking according to claim 6, characterized in that, The outer loop control signal includes angular velocity reference commands. and angle reference instructions : ， In the formula The angular velocity command is obtained from the MPC-based outer-loop vision servo controller. The angular velocity command is obtained from the outer-loop programmable logic controller based on UDE. To The optimal nominal control signal obtained through optimization; Integrating the above equation yields the angle reference command. : ； The output outer loop control signal is then... .

8. The anti-interference line-of-sight stabilization control method for dynamic target tracking according to claim 7, characterized in that, The UDE-based inner-loop attitude controller first processes the outer-loop control signal... Decomposed into pitch channel attitude commands and yaw channel attitude commands In the formula For pitch angle command, This is the yaw angle command; The nominal control input for the pitch channel of the line-of-sight stabilization device. for: ， In the formula The pitch channel inertia coefficient, For pitch channel proportional feedback gain. For the pitch channel differential feedback gain, ; For pitch angle error, ; These are the damping parameters for the pitch channel; The nominal control input for the yaw channel of the line-of-sight stabilization device. for: ， In the formula This is the inertia coefficient of the yaw channel. This is the proportional feedback gain for the yaw channel. For the differential feedback gain of the yaw channel, ; For yaw angle error, , The yaw angle of the line-of-sight stabilizing device; These are the damping parameters for the yaw channel.

9. The anti-interference line-of-sight stabilization control method for dynamic target tracking according to claim 8, characterized in that, The UDE-based inner-loop attitude controller is also used to calculate pitch channel disturbance estimates. And yaw channel interference estimates : ， ， In the formula For pitch channel UDE parameters, ; For yaw channel UDE parameters, .

10. The anti-interference line-of-sight stabilization control method for dynamic target tracking according to claim 9, characterized in that, The control signals output by the UDE-based inner-loop attitude controller for the line-of-sight stabilization device include a composite control signal from the pitch channel. Combined control signal with yaw channel : ； 。

Images